Re: plain rules, please [was: Semantic Web Rule Language (SWRL) 0.5 released]

On Sunday, November 30, 2003, at 04:58 PM, Drew McDermott wrote:
>    [Bijan Parsia]
>    ... except that
>    Shirky seems to be arguing, if against anything, against formality 
> and,
>    you in so far as I can read your computation vs. deduction 
> paragraph as
>    a half-agreement with Shirky, seem to agree, but I know that's 
> wrong.
>
> I'm in favor of formalizing that which can be formalized.  Like
> Wittgenstein.

Wittgenstein can be formalized??!!??!

:)

> I think the semantic web is going to be important in areas that have a
> reasonable expectation of being formalizable, including things like
> routine commercial transactions, certain kinds of information
> retrieval, route planning, scheduling, optimization given well-defined
> objective functions, combination of evidence sources for narrowly
> circumscribed domains, and so forth.  But even in these domains
> deduction has no special status except that it is the most
> conservative inference technique around.

So, who disagrees? In the recent debate twixt NAFers and \+NAFers, I 
didn't see anyone arguing that non-mon was *pointless*, or 
*unnecessary* (at least, in the long run). The arguments were more that 
we didn't know how to do it (right) (at least, for RDF and OWL) and we 
should, short term, stick with what we know.

Since everyone and their ferret agrees that a reasoner is going to be a 
*component* of any real world Semantic Web software or agent, I don't 
any non-choir folks to preach to (on this point). (I think there's also 
general agreement that there might be all sorts of "reasoning services" 
that express their stuff in OWL or RDF, but reason invalidly. They just 
aren't standardized by the current documnents.) It's sorta like 
Shirky's "Oh, a global, monolithic ontology is just a bad idea". Well, 
ahem, *yeah*, and, while not universal, being against such things is a 
very *strong* trend in Semantic Web circles.

Oh, to add a comment on an unrelated thread: Yay, explicit quantifiers, 
but it's not just prolog or nomalization envy. Description Logics 
generally have both implicit (e.g., for class or property subsumption 
axioms) and explicit (someValuesFrom, allValuesFrom, and the 
cardinalities) quantifiers. So does OWL. Hence, so does OWL Rules.  
This isn't fully satisfying, I agree, of course.

Cheers,
Bijan Parsia.

Received on Monday, 1 December 2003 08:31:16 UTC